The experiments presented here were designed to test a critical prediction of the colloid-osmotic hypothesis (), that parasite development is accompanied by a progressive decline in hemoglobin concentration in the host cell cytoplasm 
. We discuss first some aspects of the methodological approach applied in this investigation. As highlighted in the Introduction, the optimal way to test this hypothesis is to map the local hemoglobin concentration in the cytoplasm of intact, live, infected red blood cells. The method developed and applied for such measurements here was based on fluorescence quenching of the cell-incorporated fluorophore calcein. Both in vitro
(lysate) and in situ
(intact RBC) calibrations using the fluorophore calcein ( and ) proved that fluorescence lifetime is sensitive to the local hemoglobin concentration. The observed reductions in fluorescence lifetimes proved that molecular crowding induced FRET is indeed the main cause of quenching and that it can be applied to estimate local [Hb].
Confocal detection together with TCSPC provided the spatial resolution required for the quantitative mapping of [Hb] in the host cell cytosol. However, the limited number of photons per pixel (~300) that could be collected in 60 s of acquisition time made image binning (5×5 pixels, see ) necessary to reach the high signal-to-noise ratios required to fit multi-parameter fitting curves (eq. S2 and eq. S6 
). This standard image processing blurs structures in the sample and reduces the contrast in the cytosol of infected RBCs. The analysis of the average fluorescence lifetimes (eq. S6) indicated a significant reduction in [Hb] (data not shown), but did not provide the necessary spatial resolution. We therefore used a recent technique representing lifetime data based on AB-plots (or phasors) which map the time-resolved fluorescence decays to a linear bidimensional space where trajectories of clusters can be interpreted 
. AB-plots are particularly useful for fluorescence lifetime data segmentation 
because regions of the sample with similar fluorescence lifetimes will cluster together in the phasor space.
The results showed that the [Hb] in the host cytosol of parasitized cells decreased from values around 7.5 mM in the RBCs used for cultures, to values between 2.1–7.1 mM in IRBCs, well within the range predicted by the numerical model (). This approach has allowed us to present a more detailed view of the molecular processes occurring in a living IRBC. In fact, one of the most important advantages of this representation is the possibility of segmenting regions of the sample that exhibit different fluorescence lifetime prior to any data fitting 
. Thanks to this strategy, it was possible to discriminate clearly between parasite and RBC cytoplasm. The photons collected from the segmented regions (105
) provided the signal-to-noise ratios required to fit the proposed physical model (eq. 1) and to demonstrate the reduction in cytosolic [Hb]. The present results, documenting progressive dilution of [Hb] as the parasite matures, highlight the importance of studying live, unprocessed cell samples when evaluating homeostatic parameters. After submission of our manuscript a paper by Park and colleagues 
appeared documenting an overall decline in host haemoglobin concentration with parasite maturation similar to the one reported here, by applying a combination of tomographic phase microscopy and difraction phase microscopy. The main prediction of the colloidosmotic hypothesis is thus supported by two independent observations obtained with different methods, both on live cell samples.
A surprising finding in the phasor analysis was the presence of a component with distinctive properties in IRBCs with mature parasites. Edge effects caused by RBC motion artefacts can be excluded because healthy RBCs show, at worse, a very narrow edge defined by slightly different lifetimes, which can be attributed to the mixing of calcein lifetime with background photons, resulting in faint tails in the phasor phase (). The distinct appearance of the three clusters in phasor diagrams of trophozoite- IRBCs (see ) suggests the formation of microdomains within the host cell cytoplasmic environment, in which hemoglobin appears largely excluded from close contact with calcein. In young trophozoites these domains are seen mainly as an extensive peripheral zone beneath the IRBC plasma membrane, with a similar though narrower zone around the parasite. In more mature IRBCs where the parasite is larger, the two regions coalesce, with no intervening Hb-rich zone (). This change is paralleled by an increased compartmentalization measured for calcein.
These developments are likely to reflect localized changes induced in the IRBC by the parasite (see ), including the export and assembly of molecules destined for Maurer's clefts, knobs and other parasite-derived structures in the peripheral zone of the host cell 
. We note that also the fluorescence intensity of calcein appears higher at the cell periphery of trophozoite containing IRBCs (arrows in ), but not in uninfected RBCs () or ring-stage IRBCs (), supporting the intepretration of significant calcein de-quenching beneath the membrane surface. Likewise, the traffic of parasite proteins from the parasitophorous vacuole surrounding the trophozoite is intense at this stage 
, and this may be related to the altered region close to the parasite surface. The model parameters from which the colloidosmotic hypothesis was derived ensure maintenance or rapid restoration of osmotic equilibria between host cell cytoplasm and extracellular medium. Could the Hb-restricted microdomain under the host cell membrane interfere with such equilibria? It is well established that isotonic sorbitol or alanine retain their lytic potential of IRBCs with parasites of all stages beyond the ring stage indicating that the Hb-restricted domain under the membrane represents no limiting permeability barrier to NPP-mediated or water fluxes. Therefore, any domains between the bulk Hb-containing cytosol of the host cell and the external medium cannot alter the way in which osmotic equilibria are maintained or restored between these two compartments.
Interpretative diagram illustrating the observed [Hb] and compartmentation effects throughout parasite development.
In conclusion, the present results demonstrate that the well known reduction in Hb content of malaria-infected RBCs is accompanied by a concomitant reduction in Hb concentration, providing strong support for the colloid-osmotic interpretation of excess Hb consumption. They also reveal the existence of a submembrane compartment where Hb is partially excluded, locally diluted, or both. The molecular basis of this change awaits a more detailed analysis of the complex parasite-host cell interaction.